Composite Measure References:
Hi. The following references might be of assistance with your composite measure:

CMS Blueprint:
https://www.cms.gov/Medicare/Quality-Inivtiaties-Patient-Assessment-Instruments/MMS/Downloads/Blueprint.pdf

ECQI Resource Center links to CQL HQMF IG:
https://ecqi.healthit.gov/tool/hl7-cql-hqmf-ig

CQL HQMF IG R1 STU4 July 2019:
http://www.hl7.org/implement/standards/product_brief.cfm?product_id=405

Pertinent areas of zip file:
CQL HQMF IG R1 STU4, V3, Volume 1, Section 6: Composite Measure Development
Section 6.3: Patient-level Linear Combination Scoring
using CMPMSRMTH with LINEARSCR for continuous variable scoring

CQL HQMF IG R1 STU4 examples folder:
PatientLinearCombination_v5_4_Artifacts

Hi. The following might be of assistance with your composite measure.
https://www.cms.gov/Medicare/Quality-Inivtiaties-Patient-Assessment-Instruments/MMS/Downloads/Blueprint.pdf
https://ecqi.healthit.gov/tool/hl7-cql-hqmf-ig
http://www.hl7.org/implement/standards/product_brief.cfm?product_id=405
PMuir MD ESAC Standards Team

Thank you, I'm a bit behind the ball and am now reviewing the test measure you started up, I will likely work off of that one and complete the process and develop the testing. I appreciate your insight on how best to get to the scoring stage, I will review your feedback and return with any questions. This has been exceedingly helpful and I appreciate your patience with me as it was challenging not having too many examples to work from. 

Best,

Angel

CQLIT-227 re CQM877 Global Malnutrition Composite Score
Composite Measure Construction in MAT, request guidance on calculation with linear scoring measures

Further review confirms that the best recommendation at this point is to combine the four component measures into one proportional measure for hospital episodes.

This can be evaluated as a proportional measure by assigning each of the component measure as a distinct “Numerator One”, “Denominator One” and “Denominator Exclusions One” for Component Measure One through “Numerator Four”, “Denominator Four” and “Denominator Exclusions Four” for Component Measure Four.

A simple test configuration (TestPJMCQLIT227pr.cql) in Bonnie confirmed that each hospital episode collated the data elements correctly with stratification by “Numerator One” through “Numerator Four” (component measure). The xlsx spreadsheet output from Bonnie displays the data as a tab for each Numerator (component measure) as Numerator, Denominator and Denominator Exclusion.

Determining the score of an individual hospital episode is complicated as each stratification within the episode is calculated separately so that a blending function is challenging. This is compounded by one of the four denominators having an additional expression. However, since your equation to combine the measure populations is focused on the overall inpatient episode population rather than the individual episode, perhaps there is an alternative pathway.

Since hospital measures report their data as QRDA I which contains the appropriate data elements for each hospital episode rather than just the summarized total as in QRDA III, the receiving system could perform the calculation across all hospital episodes for all of the component measures utilizing your equations.

I hope that this information is helpful.
PMuir MD ESAC StandardsTeam

Upon further review and having found additional documentation on the requirements of the component nutrition measures, while a continuous variable measure provides functionality for the Total Composite Score, it complicates matters since the established component measures are proportional. Developed a four component proportional measure model that works appropriately in Bonnie. As an EH measure, the output would be reported in the QRDA I as discrete data for each component for each episode so that the receiving system could do the calculation. Do not have a TestPJMCQLIT227pr-0.0.009.cql totaling mechanism within the measure at this point.

I have attached TestPJM227pr.cql as example of 4 subcomponent measure. As you will see the component definitions pull data from the encounter for easy testing in Bonnie.

PMuir MD ESAC Standards Team

I have attached TestPJMCQLIT227cv.zip plus CMSv0(1).xlsx which contains continuous variable with summary of two cases, plus screen shots.
While not a tree trunk, it is at least a sapling showing you how to do the main structure with measure observation as an average aggregate function and the components as definitions, exclusions as measure exclusions, also threw in stratification as example.
You should be able to load into Bonnie (I created in MAT on FHIR and loaded into Bonnie) but I would suggest recreating so you can start to fill in the branches and leaves.
Let me know if you have any further troubles with implementing.
PMuir MD ESAC Standards Team

additional info and example re continuous variable, albeit from the CQL in FHIR
http://hl7.org/fhir/us/cqfmeasures/2020Feb/measure-conformance.html#continuous-variable-measure

see Section 3.4.5 Continuous Variable Measure
Note that CV does not have Denominator Exclusions, however, Initial Population would be same and Measure Population could be IP - Exclusions.

define "Measure Population" :
"Initial Population"

define "Inpatient Encounter" :
["Encounter"] Encounter
where LengthInDays(Encounter.period) <= 120
and Encounter.period ends during "Measurement Period"

//Measure Observation
define function "Related ED Visit" (encounter Encounter) :
Last(["Encounter" : "Emergency Department Visit"] ED
where ED.status = 'finished'
and ED.period ends 1 hour or less before start of encounter.period
sort by start of period )

define function "Measure Observation" (encounter Encounter) :
duration in minutes of "Related ED Visit"(encounter).period

Some other helpful references as background:
https://ecqi.healthit.gov/sites/default/files/Guide-for-Reading-eCQMs-v6.pdf
https://ecqi.healthit.gov/system/files/eCQM-Logic-and-Guidance-2018-0504.pdf

Atom editor, available for download from atom.io
install packages: atom-ide-ui v0.13.0, ide-java v0.9.0, language-cql v2.6.0

enable MAT CQL Editor mode in user profle

Summary of suggestions:

• remove include libraries (note for future, avoid underscores in library names),
• incorporate component measures into one measure
• remove duplicate definitions
• prefer 'during' over 'in' for timing restrictions (consider 'overlaps' if pertinent to intent)
• definition naming conventions (see Style Guide, Authoring Guide, etc)
• eg define "MN Diagnosis Denominator"
• avoid abbreviations,
• name should describe type of results returned (eg "Encounters with ...")
• this definition uses intersect - should this be union
• exists() returns true if contents are present (may replace "is not null")
• Global_Malnutrition... define Denominator Exclusions error (Exclusion?)
• suggest continuous variable with definitions
• combine Denominator Exclusions into single
• define measure and observation populations
• use measure observation 1, 2, 3, 4 for component scores
• define measure observation 5 as (1 + 2 + 3 + 4) /4
• assign measure observations in packaging
• data model and valueset discussion

Hope this is of help
PMuir MD ESAC Standards Team

Hello, I've updated the Global Composite Measure Library to include all the sub-libraries from all the component measures as you suggested above. I think where I'm struggling and would need support is how to translate the calculations to the measure packager. One point of clarification from your assumptions of how this measure is calculated, this is not a patient-level measure, it is a hospital-level measure. Each component is calculated at the HOSPITAL level then averaged out to give a composite score. "Patients" are not scored. Hope that helps as well. 

I've gone through and eliminated any inconsistencies in the common libraries across the 4 components. 

For ease of reference at this time, I've included the operations for calculating each component score both in the definitions and in the functions (can delete from where they shouldn't be as instructed) as follows:

Component 1: Count("Malnutrition At-Risk")/ Count("Malnutrition Screening")

Component 2: Count("Nutrition Assessment")/ ( Count("Assess Denominator")- Count("Denominator Exclusions"))

Component 3: Count("MN Diagnosis")/ ( Count("MN Diagnosis Denominator")- Count("Denominator Exclusions"))

Component 4: Count("Care Plan")/ ( Count("Care Plan Denominator")- Count("Denominator Exclusions"))

Total Composite Score: ( "Component 1" + "Component 2" + "Component 3" + "Component 4" ) / 4

Looking forward to discussing how best to proceed and potentially holding a call to discuss all assumptions and the best approach.

After discussion with Bryn and Floyd, we are looking into arranging a phone call to discuss this ticket with you. We will contact you once a coordinated time line is determined.

In the interim, you may obtain that MAT User Guide 5.9 of 5/2020 from the MAT website under Helpful Links.
https://www.emeasuretool.cms.gov/training-resources

Please refer to the User Guide for information on the CQL Library and CQL Composer as opposed to Measure Library and Measure Composer (see pg 29 of User Guide), but the simplest approach would be to have all the expressions contained within the one measure.

To assist our comprehension of your project, it would be helpful to know more about the 4 component assessment/procedures that are to make up the measure.

Thank you,
PMuir MD ESAC Standards Team

CQLIT-227 re CQM877 Global Malnutrition Composite Score
Composite Measure Construction in MAT, guidance on calculation with linear scoring measures

Your question provides an excellent teaching opportunity for others regarding measure construction using CQL expressions, so we will provide a more detailed solution for this ticket. Thank you for including the existing cql files for the measure.

After pondering on this ticket for a while, there is benefit to discussing measure design from the ground up so that we can cover the most important steps from the CQL developer point of view.

First step is the quality measure concept itself:

• What is the quality of care concept that we are trying to assess?
• Is this measure bound to the patient, by encounter, by provider, by facility type, or other population?
• What types of information are related to the quality measure?
• How is this information captured by multiple vendor EHR systems involved with patient care?
• How does that information correlate with data elements within the quality data model (QDM 5.5)?
  Thank you Floyd for providing QDM design guidance based upon the cql files that you included.
• How can we capture the information? Consider now with QDM and in the future with FHIR?
• How do we evaluate this information? Are there existing standardized assessment tools or guidelines?
• How do quantify and standardize the evaluation tools?
• How is the measure reported? QRDA I individual detail, QRDA III population summary as currently done. Future potential using bulk data (QRDA II) or reporting via FHIR?
• Gather the data model data elements and their specifications
• Which Measure Population approach meets the specifications developed? (discussion follows)
• Develop a skeleton framework to study the data flow through the measure populations.
• Develop a text description of the actions required to meet decision point specifications
  – Translate these descriptions into the CQL equivalents
  ...

Considering the approach to Populations within Clinical Quality Measures:
There is information available on CQM Populations on the Resource Center
https://ecqi.healthit.gov/ecqms

But, see the excellent discussion under Section 3.4.2 Measure Population Semantics on
http://hl7.org/fhir/us/cqfmeasures/measure-conformance.html#continuous-variable-measure

Proportion Measures (3.4.3), Ratio Measures (3.4.4), Continuous Variable Measure (3.4.5),
Cohort Definitions (3.4.6) and Measures with Multiple Populations (3.4.7) as well as Supplemental Data Elements and Stratification.

Please refer to the excellent summary of requirements found on the HL7 website under
Table 3-1: Measure populations based on types of measure scoring.

From your description of the measure, for each hospital episode, you are gathering a composite score gathered from four sub components that calculates down to a single score for the qualifying hospital episode.

The Continuous Variable Measure (CVM, 3.4.5) would be an avenue for you to consider. While this measure type only permits Initial Population, Measure Population and Measure Population Exclusion, it can provide individual and aggregate results in the Measure Observation. Note that your Denominator Exclusions have to be redefined as Measure Observation Exclusions.

This would require cloning, reassigning the cloned measure as CVM and restructuring of your measure in the MAT. This could be created as a skeleton draft measure with dummy sub components then packaged and tested in Bonnie. The component measure definitions could then be incorporated into the main measure, which is the easiest pathway..

If you wish to continue as included libraries, they will need significant editing to resolve the duplicated definitions/functions as well as Floyd’s recommendations (this would need to be done in the library development section of the MAT). There is only benefit to this additional effort if they will be reused in other measures.

The Measure Observation would be used for the scoring calculation (recommend just a single observation) as the total calculated for the Global Malnutrition Composite Score. Recommend having component assessments as definitions rather than functions, for easier testing and evaluation in the Bonnie tool.

I hope that this information is enough to get you to the next level. Keep in touch as needed. I will monitor #227.
PMuir MD ESAC StandardsTeam

To the comments on the "common library" issues. We are not a CMS contractor, so we have no knowledge at this moment on how to add our definitions to the "common library section of the MAT". I've reviewed the MAT User Guide and have tried to look up guidance on how to do this within the MAT but have not found how to do this. Using my intuition I included the four component measure CQL libraries in the "Includes" section of the MAT for the Global Composite Measure. Please let me know if this is what you were referencing in what you requested below:

• A measure may contain sub-libraries using the "include" expression.
  These sub-libraries must be added to the common library section of the MAT so that the packager can locate them. Your sub-libraries are not located there which caused the generated error messages.
• In addition, these sub-libraries get pulled into the main measure as can be seen in the MAT where a sub-library definition or function is shown there in the measure.
  This indicates the importance of avoiding duplication of these definitions and functions across the measure and sub-libraries. If the same name is used in multiple libraries this will cause confusion when pushed together.

Acknowledging the codes above, we will review the feedback. At first glance, there may be some noted challenges in assessment vs procedure as none of the intended use of the codes is a "procedure" but actually an "assessment". I may need to engage further with the measure steward which is also the author of the value sets. Alternatively, we may need to just reference the specific codes within these value sets since they don't all apply specifically.

Adding some comments about content on this measure, especially use of the value sets to determine the criteria for compliance:

1. Malnutrition Risk Screening (2.16.840.1.113762.1.4.1095.40) – used for [“Assessment, Performed”: “Malnutrition Risk Screening”] and includes

Appropriate for “Assessment, Performed”:

• 1 observation (malnutrition universal screening tool – probably should be looking at LOINC for an observable entity as well) - “Assessment, Performed”

Not Appropriate for “Assessment, Performed”

• 3 procedures (malnutrition screening, assessment using malnutrition universal screening tool, and screening for malnutrition) – should use “Procedure, Performed” or “Intervention, Performed”
• 1 encounter – Encounter for screening for nutritional disorder – should use “Encounter, Performed”

1. Malnutrition Assessment (16.840.1.113762.1.4.1095.29) – used for [“Assessment, Order”: “Malnutrition Assessment”] and includes

Appropriate for “Assessment, Order”:

• 1 procedure (nutritional assessment) – “Procedure, Performed”

Not Appropriate for “Assessment, Order”:

• 4 findings (nutrition assessment panel, comparative nutrition assessment standards panel, nutrition assessment narrative, unspecified severe protein-calorie malnutrition, protein-calorie malnutrition of moderate and mild degree) – these are not things that are “ordered”
  • The panels should use “Assessment, Performed”
  • The narrative and the 2 ICD10CM codes could use either “Assessment, Performed” to represent observations, or “Diagnosis” to indicate their presence as conditions on a problem list.

1. Malnutrition Screening at Risk (16.840.1.113762.1.4.1095.38) – use for result of [“Assessment, Performed”: “Malnutrition Screening”] and includes:

Appropriate for [“Assessment, Performed”: “Malnutrition Screening”] result attribute – each of these codes is a qualifier value

Thanks

Floyd Eisenberg

Hi. There appears to be a number of basic design problems in your implementation:

A measure may contain sub-libraries using the "include" expression.
These sub-libraries must be added to the common library section of the MAT so that the packager can locate them. Your sub-libraries are not located there which caused the generated error messages.

In addition, these sub-libraries get pulled into the main measure as can be seen in the MAT where a sub-library definition or function is shown there in the measure.
This indicates the importance of avoiding duplication of these definitions and functions across the measure and sub-libraries. If the same name is used in multiple libraries this will cause confusion when pushed together.

Common sub-libraries are beneficial when the same expressions would be used across multiple measures (eg Global.HospitalizationWithObservation)
If your sub-libraries are oriented towards just this measure, then it would be easier to incorporate them into the main measure as component functions. This could be more efficient although items within Denominator Exclusions would have to be worked into the appropriate expressions since this population is not consistent across the sub-libraries.

You may find it useful to look at measures 56 and 66 which utilize multiple assessment types.

I have started to shuffle the sub-libraries into a test version for evaluation but wanted to give you feedback to prevent delay.

I've attached the output of the CQL library for the Global_Malnutrition_Composite_Score.txt

^{Note that at this point we have not progressed further than referencing the component measures, we need guidance on how to piece together the component scores in order to perform an average score across the 4 component measures. Not sure how best to include the components, as functions or as definitions in the MAT so that they can be properly packaged in the measure packager. If guidance can be provided on that end we can implement it in the MAT and produce the full composite measure specification to include the calculation.}

Thank you in advance for your assistance here!

Got the ancillary files, thank you.
Can you also add your latest version of Global Malnutrition Composite Score.cql v0 please
Then Bryn can have everything in one place.
When I last looked at the Global Malnutrition Composite Score.cql, it had include libraries for the ancillary files but they were not visible in the common libraries section of the MAT which would cause a problem when trying to package.
Thank you

Hello, Attempting to answer these comments and/or questions in reverse order from what was posted here:

1. I'm not sure where there are "underscores" I've reviewed all four component measures and none of them have underscores in their names or any of the definitions. If you are referring to the component measure aliases in the MAT, I have removed the underscores there and have revised them all to conform with Pascal Case. 
2. Here are the eCQM packages for all four component measures: ScreeningMalnutritionRisk_v5_8_Artifacts.zip
3. In reference to the previous post saying that the component measures are not defined.. I was following the instructions for bringing in the component measures that I saw in the MAT guide.. Is the guidance that we have to include all of the component measures' definitions in the composite definitions list as well? If so, I can work to add those, I thought it countered the point of referencing the components in the first place. 
4. Finally, the problem again is that the measure is a linear score composite measure. The guidance for constructing that calculation and using the population workspace for this type of measure is nowhere I've been able to find. The current guidance does not apply to a component-level linear composite measure where it reflects the average of the four component scores. I was playing around with the functions but honestly could not figure out how to get it to calculate appropriately. Any guidance would be helpful or perhaps setting up a call to walk through the process.

Ok We are at a standstill here.
library Global_Malnutrition_Composite_Score version '0.0.000'.cql only contains part of the required information for analysis.
It refers to sub libraries that are not visible to us in MAT
include Appropriate_Documentation_Malnutrition_Diagnosis version '1.0.000' called Component_Measure_3
include NutritionAssessment_For_ScreeningRisk version '1.0.000' called Component_Measure_2
include Screening_Malnutrition_Risk_Admission version '1.0.000' called Component_Measure_1
include Development_Nutrition_Care_Plan version '1.0.000' called Component_Measure_4

While I can find some of these by searching the main library, instead of the common library area, it is a little confusing to be sure which ones you intend to utilize.
Also, there are naming issues with underscores etc for library names.
Please refer to page 3 of the CQL Style Library:
https://ecqi.healthit.gov/tool/cql-style-guide

Would you please provide all of the libraries that are associated with the measure so that we can follow the logic. You may submit as *.cql files since packaging is part of the problem.

Until then, we can proceed no further on our end.

Additional issue:
You are using include statements for
include Appropriate_Documentation_Malnutrition_Diagnosis version '1.0.000' called Component_Measure_3
include NutritionAssessment_For_ScreeningRisk version '1.0.000' called Component_Measure_2
include Screening_Malnutrition_Risk_Admission version '1.0.000' called Component_Measure_1
include Development_Nutrition_Care_Plan version '1.0.000' called Component_Measure_4

but these are not defined in the common library section so the packager will fail
these may need to be defined as functions
I will see how evaluation proceeds in testing.

Hi. Starting the review. I will clone as test measure to evaluate.
re v0.0.000, the CQL editor indicates it is having trouble finding components of the following definition:
define "Total Composite Score":
( "Component 1 Performance Score" + "Component 2 Performance Score" + "Component 3 Performance Score" + "Component 4 Performance Score" ) / 4

The definitions referred to in the above expression are just defined as "Component 1", etc.
define "Component 1":
Count(Component_Measure_1."Malnutrition At-Risk")/ Count(Component_Measure_1."Malnutrition Screening")

Have you tried removing the four " Performance Score" part of the references from the "Total Composite Score"?

This could should then save the CQL without errors which would permit packaging to occur.
Thought this is worth a try to prevent further delay in solving your issue.

PMuir MD ESAC Standards Team

Global Malnutrition Composite Score Draft v0.0.000 I have been unable to fully process and export it since I am having trouble with the packager.

Please clarify which version of the Global Malnutrition Composite Score in the MAT that you are requesting for review. Thank you

Thank you for submitting your question. We will review your ticket and provide a response as soon as possible.
This could be a good topic for discussion on a measure group meeting. Do you have a measure name in the MAT at this time? Someone from our team will contact you to help co-ordinate a discussion time.